ORIGINAL RESEARCH article
Front. Digit. Health
Sec. Health Informatics
Volume 7 - 2025 | doi: 10.3389/fdgth.2025.1616770
Sex and Age Estimation from Cardiac Signals Captured via Radar using Data Augmentation and Deep Learning: A Privacy Concern
Provisionally accepted- Universidad Carlos III de Madrid de Madrid, Leganés, Spain
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Introduction Electrocardiograms (ECGs) have long served as the standard method for cardiac monitoring. While ECGs are highly accurate and widely validated, they require direct skin contact, are sensitive to motion artifacts, and are not always practical for continuous or unobtrusive monitoring, limiting their generalization to real-world, dynamic environments. However, radar-based technologies offer a novel, non-invasive alternative for acquiring cardiac signals without direct contact. This improves both hygiene and patient comfort, making it especially attractive for medical applications. Despite these benefits, it may raise privacy concerns, inadvertently revealing personal attributes such as sex and age. This study investigates, for the first time, how such demographic information can be inferred from radar-acquired cardiac signals. Methods To address this question, we developed a machine learning framework to predict demographic attributes from radar-based cardiac signals. These signals were transformed into scalograms—a time-frequency representation—and then classified using a Convolutional Neural Network (CNN). Given the lack of prior studies on demographic inference from radar-based cardiac signals, the generalization capabilities of existing approaches remain untested in this context. Moreover, the small size of available datasets further limits model performance. To mitigate these issues, we applied data augmentation using a Conditional Wasserstein Generative Adversarial Network (cWGAN), which generated synthetic scalograms conditioned on class labels. Notably, there are very few prior studies on data augmentation specifically for this type of signal. This strategy aimed to enhance model accuracy and generalization by enriching the training data. Results Our experiments demonstrate that data augmentation significantly improves model performance. The trained model achieved an accuracy of 78.40% in predicting the sex of individuals and 72.83% accuracy in classifying them into two age groups (18–29 and 30–65 years), despite the dataset being limited to only 30 subjects. Discussion These findings reveal a potential privacy risk associated with radar-based biometric systems. The ability to infer sensitive demographic information from physiological signals could have serious implications, particularly in secure applications such as electronic passports (e-passports), where access to RFID chip data often depends on such personal attributes. Therefore, while radar technologies offer promising advantages, their deployment must consider and address the associated privacy challenges.
Keywords: Data augmentation, deep learning, Radar-based cardiac signal, Privacy, Sex and age estimation
Received: 23 Apr 2025; Accepted: 01 Oct 2025.
Copyright: © 2025 Foronda-Pascual, Camara and Peris-Lopez. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Daniel Foronda-Pascual, dforonda@pa.uc3m.es
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.